Platform-Independent Classification System to Predict Molecular Subtypes of High-Grade Serous Ovarian Carcinoma

Deep multi-omics integration by learning correlation-maximizing representation identifies prognostically stratified cancer subtypes

Motivation

Molecular subtyping by integrative modeling of multi-omics and clinical data can help the identification of robust and clinically actionable disease subgroups; an essential step in developing precision medicine approaches.

Results

We developed a novel outcome-guided molecular subgrouping framework, called Deep Multi-Omics Integrative Subtyping by Maximizing Correlation (DeepMOIS-MC), for integrative learning from multi-omics data by maximizing correlation between all input -omics views. DeepMOIS-MC consists of two parts: clustering and classification. In the clustering part, the preprocessed high-dimensional multi-omics views are input into two-layer fully connected neural networks. The outputs of individual networks are subjected to Generalized Canonical Correlation Analysis loss to learn the shared representation. Next, the learned representation is filtered by a regression model to select features that are related to a covariate clinical variable, for example, a survival/outcome. The filtered features are used for clustering to determine the optimal cluster assignments. In the classification stage, the original feature matrix of one of the -omics view is scaled and discretized based on equal frequency binning, and then subjected to feature selection using RandomForest. Using these selected features, classification models (for example, XGBoost model) are built to predict the molecular subgroups that were identified at clustering stage. We applied DeepMOIS-MC on lung and liver cancers, using TCGA datasets. In comparative analysis, we found that DeepMOIS-MC outperformed traditional approaches in patient stratification. Finally, we validated the robustness and generalizability of the classification models on independent datasets. We anticipate that the DeepMOIS-MC can be adopted to many multi-omics integrative analyses tasks.

Availability and implementation

Source codes for PyTorch implementation of DGCCA and other DeepMOIS-MC modules are available at GitHub (https://github.com/duttaprat/DeepMOIS-MC).

Supplementary information

Supplementary data are available at Bioinformatics Advances online.

Rapid, High-Throughput Single-Cell Multiplex In Situ Tagging (MIST) Analysis of Immunological Disease with Machine Learning

The cascade of immune responses involves activation of diverse immune cells and release of a large amount of cytokines, which leads to either normal, balanced inflammation or hyperinflammatory responses and even organ damage by sepsis. Conventional diagnosis of immunological disorders based on multiple cytokines in the blood serum has varied accuracy, and it is difficult to distinguish normal inflammation from sepsis. Herein, we present an approach to detect immunological disorders through rapid, ultrahigh-multiplex analysis of T cells using single-cell multiplex in situ tagging (scMIST) technology. scMIST permits simultaneous detection of 46 markers and cytokines from single cells without the assistance of special instruments. A cecal ligation and puncture sepsis model was built to supply T cells from two groups of mice that survived the surgery or died after 1 day. The scMIST assays have captured the T cell features and the dynamics over the course of recovery. Compared with cytokines in the peripheral blood, T cell markers show different dynamics and cytokine levels. We have applied a random forest machine learning model to single T cells from two groups of mice. Through training, the model has been able to predict the group of mice through T cell classification and majority rule with 94% accuracy. Our approach pioneers the direction of single-cell omics and could be widely applicable to human diseases.

Poor-prognosis molecular subtypes in adenocarcinomas of pancreato-biliary and gynecological origin: A systematic review

Pancreato-biliary and gynecological adenocarcinomas need better tools to predict clinical outcome. Potential prognostic mesenchymal(-like) transcriptome-based subtypes have been identified in these cancers. In this systematic review, we include studies into molecular subtyping and summarize biological and clinical features of the subtypes within and across sites of origin, searching for suggestions to improve classification and prognostication. PubMed and Embase were searched for original research articles describing potential mesenchymal(-like) mRNA-based subtypes in pancreato-biliary or gynecological adenocarcinomas. Studies limited to supervised clustering were excluded. Fourty-four studies, discussing cholangiocarcinomas, gallbladder, ampullary, pancreatic, ovarian, and endometrial adenocarcinomas were included. There was overlap in molecular and clinical features in mesenchymal(-like) subtypes across all adenocarcinomas. Approaches including microdissection were more likely to identify prognosis-associating subtypes. To conclude, molecular subtypes in pancreato-biliary and gynecological adenocarcinomas share biological and clinical characteristics. Furthermore, separation of stromal and epithelial signals should be applied in future studies into biliary and gynecological adenocarcinomas.

Differential Expression in the Tumor Microenvironment of mRNAs Closely Associated with Colorectal Cancer Metastasis

BACKGROUND

Metastasis of colorectal cancer (CRC) is a major cause of CRC-related mortality. However, the detailed molecular mechanism of CRC metastasis remains unknown. A recent study showed that the tumor microenvironment, which includes cancer cells and the surrounding stromal cells, plays a major role in tumor invasion and metastasis. Identification of altered messenger RNA (mRNA) expression in the tumor microenvironment is essential to elucidation of the mechanisms responsible for tumor progression. This study investigated the mRNA expression of genes closely associated with metastatic CRC compared with non-metastatic CRC.

METHODS

The samples examined were divided into cancer tissue and isolated cancer stromal tissue. The study examined altered mRNA expression in the cancer tissues using The Cancer Genome Atlas (TCGA) (377cases) and in 17 stromal tissues obtained from our laboratory via stromal isolation using an array-based analysis. In addition, 259 patients with CRC were enrolled to identify the association of the candidate markers identified with the prognosis of patients with stage 2 or 3 CRC. The study examined the enriched pathways identified by gene set enrichment analysis (GSEA) based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) module in both the TCGA dataset and isolated stromal tissue.

RESULTS

As a result, whereas tenascin-C, secreted phosphoprotein 1 and laminin were expressed in metastatic CRC cells, olfactory receptors (ORs) 11H1 and OR11H4 were expressed in stromal tissue cells isolated from metastatic CRC cases. Finally, upregulated expression of tenascin-C and OR11H4 was correlated with the outcome for CRC patients.

CONCLUSION

The authors suggest that upregulated expression levels of tenascin-C and OR11H1 play an important role in CRC progression.

Molecular Management of High-Grade Serous Ovarian Carcinoma

High-grade serous ovarian carcinoma (HGSOC) represents the most common form of epithelial ovarian carcinoma. The absence of specific symptoms leads to late-stage diagnosis, making HGSOC one of the gynecological cancers with the worst prognosis. The cellular origin of HGSOC and the role of reproductive hormones, genetic traits (such as alterations in P53 and DNA-repair mechanisms), chromosomal instability, or dysregulation of crucial signaling pathways have been considered when evaluating prognosis and response to therapy in HGSOC patients. However, the detection of HGSOC is still based on traditional methods such as carbohydrate antigen 125 (CA125) detection and ultrasound, and the combined use of these methods has yet to support significant reductions in overall mortality rates. The current paradigm for HGSOC management has moved towards early diagnosis via the non-invasive detection of molecular markers through liquid biopsies. This review presents an integrated view of the relevant cellular and molecular aspects involved in the etiopathogenesis of HGSOC and brings together studies that consider new horizons for the possible early detection of this gynecological cancer.

The Immune Subtypes and Landscape of Advanced-Stage Ovarian Cancer

Immunotherapy has played a significant role in the treatment of a variety of hematological and solid tumors, but its application in ovarian cancer (OC) remains unclear. This study aimed to identify immune subtypes of OC and delineate an immune landscape for selecting suitable patients for immunotherapy, thereby providing potent therapeutic targets for immunotherapy drug development. Three immune subtypes (IS1–IS3) with distinctive molecular, cellular, and clinical characteristics were identified from the TCGA and GSE32062 cohorts. Compared to IS1, IS3 has a better prognosis and exhibits an immunological “hot”. IS3, in contrast, exhibits an immunological “cold” and has a worse prognosis in OC patients. Moreover, gene mutations, immune modulators, CA125, CA199, and HE4 expression, along with sensitivity either to immunotherapy or chemotherapy, were significantly different among the three immune subtypes. The OC immune landscape was highly heterogeneous between individual patients. Poor prognosis was correlated with low expression of the hub genes CD2, CD3D, and CD3E, which could act not only as biomarkers for predicting prognosis, but also as potential immunotherapy targets. Our study elucidates the immunotyping and molecular characteristics of the immune microenvironment in OC, which could provide an effective immunotherapy stratification method for optimally selecting patients, and also has clinical significance for the development of new immunotherapy as well as rational combination strategies for the treatment of OC patients.

Gene Expression Subtyping Reveals Immune alterations:TCGA Database for Prognosis in Ovarian Serous Cystadenocarcinoma

Serous ovarian cancer is the most common and primary death type in ovarian cancer. In recent studies, tumor microenvironment and tumor immune infiltration significantly affect the prognosis of ovarian cancer. This study analyzed the four gene expression types of ovarian cancer in TCGA database to extract differentially expressed genes and verify the prognostic significance. Meanwhile, functional enrichment and protein interaction network analysis exposed that these genes were related to immune response and immune infiltration. Subsequently, we proved these prognostic genes in an independent data set from the GEO database. Finally, multivariate cox regression analysis revealed the prognostic significance of TAP1 and CXCL13. The genetic alteration and interaction network of these two genes were shown. Then, we established a nomogram model related to the two genes and clinical risk factors. This model performed well in Calibration plot and Decision Curve Analysis. In conclusion, we have obtained a list of genes related to the immune microenvironment with a better prognosis for serous ovarian cancer, and based on this, we have tried to establish a clinical prognosis model.

Integration of histopathological images and multi-dimensional omics analyses predicts molecular features and prognosis in high-grade serous ovarian cancer

OBJECTIVE

This study used histopathological image features to predict molecular features, and combined with multi-dimensional omics data to predict overall survival (OS) in high-grade serous ovarian cancer (HGSOC).

METHODS

Patients from The Cancer Genome Atlas (TCGA) were distributed into training set (n = 115) and test set (n = 114). In addition, we collected tissue microarrays of 92 patients as an external validation set. Quantitative features were extracted from histopathological images using CellProfiler, and utilized to establish prediction models by machine learning methods in training set. The prediction performance was assessed in test set and validation set.

RESULTS

The prediction models were able to identify BRCA1 mutation (AUC = 0.952), BRCA2 mutation (AUC = 0.912), microsatellite instability-high (AUC = 0.919), microsatellite stable (AUC = 0.924), and molecular subtypes: proliferative (AUC = 0.961), differentiated (AUC = 0.952), immunoreactive (AUC = 0.941), mesenchymal (AUC = 0.918) in test set. The prognostic model based on histopathological image features could predict OS in test set (5-year AUC = 0.825) and validation set (5-year AUC = 0.703). We next explored the integrative prognostic models of image features, genomics, transcriptomics and proteomics. In test set, the models combining two omics had higher prediction accuracy, such as image features and genomics (5-year AUC = 0.834). The multi-omics model including all features showed the best prediction performance (5-year AUC = 0.911). According to risk score of multi-omics model, the high-risk and low-risk groups had significant survival differences (HR = 18.23, p < 0.001).

CONCLUSIONS

These results indicated the potential ability of histopathological image features to predict above molecular features and survival risk of HGSOC patients. The integration of image features and multi-omics data may improve prognosis prediction in HGSOC patients.

Molecular Oncology of Gynecologic Tumors

Oncologists and pathologists alike have recognized that the broad histologic categories, especially for ovarian and endometrial carcinomas, do not reliably segregate groups with similar clinical courses or responses to therapeutic interventions. During the last decade a paradigm shift was invoked when the results from The Cancer Genome Atlas (TCGA) project were published. Comprehensive genomic profiling data from TCGA has shown that there are four molecular subgroups of endometrioid carcinomas instead of the two subtypes proposed by Bokhman in the 1970s. For ovarian carcinomas (OC) it is now evident that molecular parameters are also significant. Although traditionally referred to as a single entity, OC is not a homogeneous disease but rather a group of diseases, each with different morphology and biologic behavior. Similar to endometrial cancers, advanced cervical cancer and recurrent disease remain particularly problematic due to chemotherapy resistance. Effective prophylactic vaccines against the most important carcinogenic human papillomaviruses (HPV) types are available, but uptake remains poor. The E6 and E7 oncoproteins are attractive targets for cancer therapy. They are constitutively expressed in HPV-positive tumors, specific to the tumor, functionally important to the tumor cells and recognized by the adaptive immune system as tumor antigens. This review summarizes recent advances in the molecular pathology, which have greatly improved our understanding of the biology of gynecologic cancers.

Are Epithelial Ovarian Cancers of the Mesenchymal Subtype Actually Intraperitoneal Metastases to the Ovary?

Primary ovarian high-grade serous carcinoma (HGSC) has been classified into 4 molecular subtypes: Immunoreactive, Proliferative, Differentiated, and Mesenchymal (Mes), of which the Mes subtype (Mes-HGSC) is associated with the worst clinical outcomes. We propose that Mes-HGSC comprise clusters of cancer and associated stromal cells that detached from tumors in the upper abdomen/omentum and disseminated in the peritoneal cavity, including to the ovary. Using comparative analyses of multiple transcriptomic data sets, we provide the following evidence that the phenotype of Mes-HGSC matches the phenotype of tumors in the upper abdomen/omentum: (1) irrespective of the primary ovarian HGSC molecular subtype, matched upper abdominal/omental metastases were typically of the Mes subtype, (2) the Mes subtype was present at the ovarian site only in patients with concurrent upper abdominal/omental metastases and not in those with HGSC confined to the ovary, and (3) ovarian Mes-HGSC had an expression profile characteristic of stromal cells in the upper abdominal/omental metastases. We suggest that ovarian Mes-HGSC signifies advanced intraperitoneal tumor dissemination to the ovary rather than a subtype of primary ovarian HGSC. This is consistent with the presence of upper abdominal/omental disease, suboptimal debulking, and worst survival previously reported in patients with ovarian Mes-HGSC compared to other molecular subtypes.

Molecular classification and immunologic characteristics of immunoreactive high‐grade serous ovarian cancer

High‐grade serous ovarian cancer (HGS‐OvCa) is one of the most lethal gynaecological malignancies. Molecular classification identified an immunoreactive subtype of HGS‐OvCa; however, the immunologic characteristics of immunoreactive HGS‐OvcA remain unclear. In this study, 121 immunoreactive HGS‐OvCa samples were identified from a meta‐analysis of 5 large transcriptome profiling data sets using a cross‐platform immunoreactive HGS‐OvCa subgroup‐specific classifier. By comparing the gene expression profiles of immunoreactive HGS‐OvCa samples and normal tissues, 653 differentially expressed genes (DEGs) were identified. KEGG pathway analysis revealed that the leukocyte transendothelial migration pathways were significantly enriched in the immunoreactive HGS‐OvCa. Protein‐protein interaction analysis identified a module that showed strong involvement of the immune‐related chemokine signalling pathway. Moreover, the GSEA enrichment analysis showed a T‐cell subgroup and M1 macrophages were significantly enriched in immunoreactive OvCa compared with normal samples. Macrophage infiltration levels were significantly elevated in immunoreactive HGS‐OvCa compared with other OvCa subtypes. In addition, expression of immune checkpoint molecules VTCN1 and IDO1 was significantly increased in immunoreactive HGS‐OvCa. In summary, our results suggest that the immunoreactive HGS‐OvCa has unique molecular characteristics and a tumour‐associated immune microenvironment featured by increased infiltration of macrophages, rather than lymphocytes. VTCN1 could be potential targets for the treatment of immunoreactive HGS‐OvCa.

Murine Oviductal High-Grade Serous Carcinomas Mirror the Genomic Alterations, Gene Expression Profiles, and Immune Microenvironment of Their Human Counterparts

Robust preclinical models of ovarian high-grade serous carcinoma (HGSC) are needed to advance our understanding of HGSC pathogenesis and to test novel strategies aimed at improving clinical outcomes for women with the disease. Genetically engineered mouse models of HGSC recapitulating the likely cell of origin (fallopian tube), underlying genetic defects, histology, and biologic behavior of human HGSCs have been developed. However, the degree to which the mouse tumors acquire the somatic genomic changes, gene expression profiles, and immune microenvironment that characterize human HGSCs remains unclear. We used integrated molecular characterization of oviductal HGSCs arising in the context of Brca1, Trp53, Rb1, and Nf1 (BPRN) inactivation to determine whether the mouse tumors recapitulate human HGSCs across multiple domains of molecular features. Targeted DNA sequencing showed the mouse BPRN tumors, but not endometrioid carcinoma-like tumors based on different genetic defects (e.g., Apc and Pten), acquire somatic mutations and widespread copy number alterations similar to those observed in human HGSCs. RNA sequencing showed the mouse HGSCs most closely resemble the so-called immunoreactive and mesenchymal subsets of human HGSCs. A combined immuno-genomic analysis demonstrated the immune microenvironment of BPRN tumors models key aspects of tumor-immune dynamics in the immunoreactive and mesenchymal subtypes of human HGSC, with enrichment of immunosuppressive cell subsets such as myeloid-derived suppressor cells and regulatory T cells. The findings further validate the BPRN model as a robust preclinical experimental platform to address current barriers to improved prevention, diagnosis, and treatment of this often lethal cancer. SIGNIFICANCE: The acquired gene mutations, broad genomic alterations, and gene expression and immune cell-tumor axis changes in a mouse model of oviductal serous carcinoma closely mirror those of human tubo-ovarian high-grade serous carcinoma.